Krashen's Five Hypotheses, in Plain English

Stephen Krashen is probably the most influential — and most controversial — figure in modern linguistics. In the late 1970s and early 1980s, he proposed a theory of language acquisition that contradicted almost everything language teaching institutions believed. He was right. Forty years of research, plus the unexpected arrival of large language models, have made his central claims hard to argue with.

Here's what he actually said, without the academic wrapper.

Hypothesis 1: The Acquisition-Learning Distinction

You acquire language one way. You learn it another. They're not the same thing.

Acquisition is the unconscious process by which you internalize a language through exposure and use. It's how you got your first language. You weren't taught rules — you were immersed in examples until the patterns became automatic.

Learning is the conscious study of language — grammar rules, vocabulary lists, conjugation tables. You know you've learned something when you can explain it. You know you've acquired it when you just do it without thinking.

The key Krashen claim: learned language and acquired language are stored differently in the brain, serve different functions, and cannot be converted into each other. Studying grammar does not eventually become acquired fluency. It stays in the monitor (more on that below) and never crosses over.

This is the most provocative of his claims and the one that generates the most resistance from traditional language teachers, because it implies that most of what happens in a classroom doesn't produce the thing learners want.

Hypothesis 2: The Natural Order Hypothesis

Languages are acquired in a predictable sequence, regardless of the order they're taught.

Across every language that's been studied, and across every learner regardless of their native tongue, certain structures are acquired before others. In English, for example, learners consistently acquire progressive tense (-ing) before third-person singular present tense (-s). The order is remarkably stable across learner groups.

The implication: you can teach grammatical structures in any sequence you like, but learners will only acquire them when they're ready. Teaching the subjunctive in Week 3 doesn't mean anyone will acquire it in Week 3. The brain has its own queue.

This is why drilling a grammatical structure doesn't mean you've acquired it. You may be able to produce it correctly in a controlled exercise and still not have it available when you need it in conversation.

Hypothesis 3: The Input Hypothesis (i+1)

You acquire language when you understand messages slightly beyond your current level.

This is the central claim. Krashen called it "comprehensible input" — input in the zone where you can understand the message even if you don't know every word. The formula is i+1: your current level (i) plus a bit more (+1).

When input is in this zone, something happens below conscious awareness: the brain extracts patterns from the data. Grammar is not taught — it's inferred from thousands of examples. Vocabulary is not memorized — it's learned from context, across many encounters.

This is exactly how large language models work. GPT was not taught grammar rules. It was trained on hundreds of billions of words of human language until the patterns became part of its architecture. Same mechanism, different implementation.

The practical implications:

Content that's too easy doesn't push acquisition forward
Content that's too hard doesn't either (you can't infer patterns you can't see)
The sweet spot is mostly-understandable with some stretch
Repetition across different contexts accelerates acquisition

Hypothesis 4: The Monitor Hypothesis

Learned grammar has one narrow job: checking output when you have time and attention to spare.

The monitor is your conscious grammar brain. It's the part of you that knows the rule for agreement and can apply it when you're writing slowly and carefully. It kicks in during proofreading. It occasionally fires mid-sentence when you catch yourself about to make a mistake you know is a mistake.

That's it. That's the monitor's entire domain.

In real conversation — the kind that happens in real time, with a real person, with social pressure and limited processing bandwidth — the monitor doesn't have time to engage. Fluent production in real time comes entirely from acquired language. The monitor can tidy up the edges afterward; it can't run the engine.

This is why you can know a grammar rule perfectly and still get it wrong in conversation. And why people who've never studied grammar can speak it correctly — because they're drawing on acquisition, not rules.

Hypothesis 5: The Affective Filter Hypothesis

Emotional state determines whether input can be acquired.

The affective filter is Krashen's term for the mental and emotional barrier that controls how much input reaches the acquisition mechanism. When the filter is low — when you're relaxed, curious, motivated, not anxious about performance — input flows through. When the filter is high — when you're stressed, self-conscious, afraid of mistakes, under pressure to perform — input stops landing.

This is why:

Anxiety about speaking early actively harms acquisition
Classrooms that emphasize error correction raise students' filters
People who "can't learn languages" often turn out to be people in high-pressure learning environments
You can listen to hours of input in a bad state of mind and acquire almost nothing

It's also why the emotional design of a language learning environment matters. Boring, stressful, or humiliating input environments don't just feel bad — they biologically block the process they're supposed to trigger.

Why This All Holds Up

Krashen's framework made specific, testable predictions. Researchers have spent forty years trying to find the holes. Some details have been refined. The core has held.

The clearest modern validation came from an unexpected direction: AI. Large language models acquire language — all languages, simultaneously — through nothing but comprehensible input at massive scale. No grammar instruction. No drills. No correction. Just exposure to meaning-bearing language until the patterns internalize.

If you wanted to build a system that maximizes language acquisition based on Krashen's framework, it would look very much like what we built.